A new open-source Bitcoin seed generator leverages radioactive decay events to create BIP39 recovery phrases, available in both 12-word and 24-word formats, using an offline hardware device.

The Entropy32 Plus project, initiated by developer Alexander Higgins in September 2026, offers a repository filled with firmware, circuit designs, fabrication documentation, and printable enclosures. Higgins emphasizes that the device is intended for experimental educational purposes and advises users to refrain from entrusting it with significant funds until a thorough independent evaluation is conducted.

Dubbed “The Universe Bifurcator,” this device captures pulses from a Geiger counter, measuring intervals between detected decay events. The firmware translates the comparison of these consecutive time intervals into bits, creating a 512-bit pool which is then processed using SHA-256.

The subsequent hash generated by Entropy32 Plus is formatted into a conventional BIP39 recovery phrase. Notably, it operates independently of Wi-Fi, Bluetooth, or any persistent seed storage per its design documentation.

A Bitcoin Seed Generator Is Using Radioactive Decay for Entropy

An open-source venture named Entropy32 Plus harnesses the randomness from radioactive decay to produce Bitcoin BIP39 seed phrases. A Geiger counter captures decay events, which are processed through… pic.twitter.com/ZhREDr89HQ

— Wu Blockchain (@WuBlockchain) September 13, 2026

Entropy32 Plus utilizes timing of radiation events

Radioactive decay happens randomly at the atomic level. Instead of relying purely on software-generated randomness, Entropy32 Plus utilizes the timing of decay events as its physical randomness foundation.

The current version accepts pulses via a 3.5-millimeter connection from a GQ Electronics GMC-320S Geiger counter, which can detect emissions from various radioactive sources or ambient radiation.

Upon receiving a pulse, an LM393 comparator converts the analog signal into a digital input, which is then processed by an ATmega328P microcontroller.

This project’s firmware logs each pulse event with microsecond timing. It contrasts the latest interval with the prior one, generating a ‘one’ if the latest interval exceeds the previous one and a ‘zero’ otherwise.

Intervals that are equal are disregarded, and any measurement shorter than 200 microseconds is considered a possible glitch, while accepted comparisons continue until 512 bits are gathered.

Higgins notes that this method of using pulse timing is significant because predicting the exact timing of individual decay events is impossible. However, he does not assert that each collected comparison yields a full bit of measured entropy.

SHA-256 processes the raw Bitcoin seed data

Once the 512-bit pool is established, Entropy32 Plus processes the complete 64 bytes using its integrated SHA-256 algorithm. The firmware employs the first 16 bytes of the resulting hash for a 12-word phrase, or all 32 bytes for a 24-word phrase.

This hashing function aims to obscure any discernible patterns related to detector dead time, count-rate fluctuations, or correlations between measurements. While hash conditioning can distribute entropy across an output, it doesn’t guarantee how much unpredictability is present in the original data.

The firmware includes a startup check for SHA-256 by hashing the string “abc” and comparing it with a known output. If this check fails, the device halts before gathering any entropy.

This self-testing ensures that the implementation produces the expected results for the designated test vector, albeit without verifying the Geiger counter’s accuracy, pulse independence, or quantifying the minimum entropy fed into the hash.

The project collects twice as many comparison bits as the maximum 256-bit BIP39 input it generates. Without formal measurement of the physical source, the repository does not specify how much usable entropy each comparison yields.

The device maps processed data to BIP39 words

Following SHA-256 processing, the device applies BIP39 standards, converting the approved entropy length into a mnemonic sentence sourced from a standardized list of 2,048 words.

For generating a 12-word phrase, Entropy32 Plus utilizes 128 bits of processed data and calculates a four-bit checksum. The cumulative 132 bits are segmented into twelve 11-bit groups, each used to select a word.

In the 24-word setting, it employs 256 bits along with an eight-bit checksum, dividing the resulting 264-bit string to produce 24 word indexes.

A seed phrase is essential for restoring wallet access, as detailed in Crypto.news’ guide. Anyone with the correct sequence of words can regenerate the wallet’s keys.

It’s important to note that a seed phrase is not merely a collection of randomly generated words. As explained in Crypto.news discussions of public and private keys, contemporary Bitcoin wallets derive multiple private keys from one master seed. Thus, any weaknesses in the initial entropy can potentially jeopardize every address derived from it.

The Entropy32 Plus firmware incorporates the complete BIP39 English word list, taking up 13,117 bytes, which is roughly 43% of the microcontroller’s flash memory capacity.

The compiled code, which includes the OLED interface and SHA-256 logic, utilizes 30,006 out of 30,720 available bytes on the chip, leaving just 714 bytes unused.

Offline design ensures seed safety from connected devices

The hardware files released do not signify any wireless communication capabilities. Users interact with the generator via two buttons, while the generated phrase is displayed on a 0.91-inch OLED screen.

Generated word indexes briefly remain in the microcontroller’s RAM. After the user records the phrase, pressing both buttons initiates a confirmation process that overwrites the word array and restarts entropy collection.

The firmware discards the original pool after it has been processed. Temporary buffers for hashes, checksums, and entropy are cleared once the BIP39 indexes have been generated.

As per the developer, the device does not retain a recovery phrase in non-volatile memory. The open-source design lacks a secure element or any mechanism to verify that the installed firmware matches the publicly available code.

The physical isolation reduces attack vectors associated with network access but does not guarantee proper assembly or trustworthiness of the software loaded. A modified device might display predetermined words or retain generated data without showing any visible differences.

According to Crypto.news’ wallet security guidelines, it’s advisable to keep recovery phrases separate from phones, cloud storage, photographs, and other interconnected systems. The device’s screen-based process aligns with offline transcription, given that the hardware and firmware are deemed secure.

Independent entropy testing is still pending

The repository indicates that the entropy derived from the Geiger source has yet to undergo evaluation per NIST Special Publication 800-90B standards, which detail the assessment of physical entropy sources and outline distinct methods for analyzing independent versus non-independent data.

Higgins encourages collecting raw inter-arrival times over an extended duration and analyzing them with NIST’s non-IID minimum-entropy estimators. The firmware comments caution that relying solely on general statistical tests is insufficient to validate adequate entropy.

“Please verify the entropy source you plan to use; otherwise, proceed with the understanding that usage is at your own risk,” the README advises.

The developer further suggests combining the Geiger-derived data with another independently validated source prior to creating a wallet for significant funds. No external audits, NIST evaluations, or public raw-event datasets were available in the repository as of September 13.

Previous wallet failures underscore the need for thorough entropy testing beyond simply confirming that the final output adheres to BIP39 formatting. As reported by Crypto.news, Kraken’s security head advocated for independent wallet assessments following vulnerabilities discovered in Coldcard seed generation processes.

A related technical analysis highlighted how weak entropy could imperil Bitcoin wallets without revealing user identities. Attackers can create potential seeds, generate corresponding public addresses, and scan the blockchain for matches.

Crypto.news further reported that issues with predictable seed generation persisted exploitable vulnerabilities for years, demonstrating the importance of rectifying faulty firmware; users must create new wallets and transfer assets if earlier vulnerabilities were exploited.

Entropy32 Plus continues to evolve actively, with its commit history reflecting an initial update from September 4, followed by modifications to firmware, circuit designs, and enclosures. There were no official production releases or timelines for completed independent security assessments as of September 13.

Share.